ORIGINAL CONTRIBUTION

A Cross-sectional Study of Emergency

Department Boarding Practices in the United
States
Stephen R. Pitts, MD, MPH, Frances L. Vaughns, PhD, Marc A. Gautreau, MD, MBA, Matthew W.
Cogdell, MS, MPH, and Zachary Meisel, MD, MPH, MSc

Abstract
Objectives: The median emergency department (ED) boarding time for admitted patients has been a
nationally reportable core measure that now also affects ED accreditation and reimbursement. However,
no direct national probability samples of ED boarding data have been available to guide this policy until
now. The authors studied new National Hospital Ambulatory Medical Care Survey (NHAMCS) survey
items to establish baseline values, to generate hypotheses for future research, and to help improve survey
quality in the future.
Methods: This was a cross-sectional, multistage, stratified annual analysis of EDs and ED visits from the
National Hospital Ambulatory Medical Care Survey public use files from 2007 to 2010, a total of 139,502
visit records. These data represent the only national measure of ED boarding. The main outcome of
interest was boarding duration for individual patient visits. Data analyses accounted for complex
sampling design.
Results: The national median boarding time was 79 minutes, with an interquartile range of 36 to 145
minutes. The prevalence of boarding for more than 2 hours among admitted patients was 32% (95%
confidence interval [CI] = 30% to 35%). Average ED volume, occupancy, acuity, and hospital admission
rates increased abruptly from the second to the third quartile of median boarding duration. The half of
hospitals with the longest median boarding times accounted for 73% of ED visits and 79% of ED
hospitalizations nationally. Thirty-nine percent of EDs (95% CI = 32% to 46%) reported never holding
patients for more than 2 hours, but visit-level analysis at these EDs found that 21% of admissions did in
fact stay in the ED over 2 hours. Only 19% of EDs (95% CI = 16% to 22%) used a strategy of moving
admitted patients to alternative sites in the hospital during crowded times.
Conclusions: In this national survey, ED boarding of admitted patients disproportionately affects
hospitals with higher ED volumes, which also see sicker patients who wait longer to be seen, but not
hospitals with higher proportions of Medicaid or uninsured visits. This finding implies that, unlike other
quality measures, there is a negative volume-outcome relationship for timely hospitalization from the ED.
ACADEMIC EMERGENCY MEDICINE 2014;21:497–503 © 2014 by the Society for Academic Emergency
Medicine

I
ncreasing emergency department (ED) visits have
caused rising apprehension in the medical commu-
nity for over half a century,1 and the trend in ED
crowding continues unabated.2 A 2009 U.S. Govern-
ment Accountability Office (GAO) report concluded that
the most important cause of ED crowding is the lack of
access to inpatient beds, leading to prolonged ED stays
for admitted patients, often in ED hallways.3 These
patients are considered to be “boarding” in the ED.
Since the GAO report, there have been several

From the Department of Emergency Medicine, Emory University School of Medicine (SRP), Atlanta, GA; the Office of the Assis-
tant Secretary for Preparedness and Response, Emergency Care Coordination Center (FLV, MWC), Washington, DC; and the
Department of Emergency Medicine, University of Massachusetts Medical School (MAG), Boston, MA.
Received August 26, 2013; revisions received October 26 and November 13, 2013; accepted November 19, 2013.
During the work for this study Dr. Pitts’ salary was funded in part by an IPA contract from the Department of Health and Human
Services. The rest of the authors have no relevant financial information or potential conflicts of interest to disclose. Dr. Meisel, an
associate editor for this journal, had no role in the peer review process or publication decision for this paper.
A related article appears on page 570.
Address for correspondence and reprints: Stephen R. Pitts MD, MPH; e-mail: srpitts@emory.edu.

© 2014 by the Society for Academic Emergency Medicine ISSN 1069-6563 497
doi: 10.1111/acem.12375 PII ISSN 1069-6563583 497
498
empirical, geographically limited studies of ED board-
ing.4–7 One 2008 study inferred boarding times indi-
rectly from data on overall visit length in the 2003
through 2005 national ED surveys.8 The GAO report
noted that while representative visit-level measures of
boarding were being collected by the National Center
for Health Statistics (NCHS), a branch of the Centers
for Disease Control and Prevention, they were not yet
available at that time. Nevertheless, after consensus
meetings of key stakeholders organized by the National
Quality Forum, the Centers for Medicare and Medicaid
Services, and the Joint Commission, the Joint Commis-
sion adopted median boarding duration as a publicly
reported core measure; these data are available through
accessing the Centers for Medicare and Medicaid Ser-
vices website “Hospital Compare”9 and began affecting
Medicare reimbursement in 2014.
Because the ED is increasingly the site of acute care
for serious medical problems,10 and has become the
main portal of entry for unscheduled hospital admis-
sions,11 the boarding problem has become increasingly
relevant to national policy, reflecting inefficiency or lack
of capacity in the overall health care system. Our goal is
to evaluate the size and character of this problem by
analyzing the first national survey of boarding items in
the National Hospital Ambulatory Medical Care Survey
(NHAMCS), the only national probability survey of ED
visits. We hypothesized that national data would reveal
a widespread distribution of boarding in the United
States and that boarding would be associated with ED
characteristics unrelated to ED performance, such as
demographic composition and the volume of visits.
METHODS
Study Design
This was a cross-sectional analysis of a nationally repre-
sentative sample of ED visits. The NHAMCS protocol
has been approved by the NCHS Research Ethics
Review Board annually since February 2003. Initiated in
1992, the NHAMCS abstracts approximately 35,000 to
40,000 patient record forms from 350 to 400 EDs annu-
ally. The survey is described in detail on the NCHS
website.12
Study Setting and Population
Some ED operational metrics, including ED-specific
annual volumes, could not be included in a model with
the ED as unit of analysis because ED-specific values
are excluded from public-use data by the NCHS, to min-
imize the risk of disclosure of hospital identity. Since
these characteristics could not be included in a multi-
variable model, we aggregated them within quartiles of
median boarding duration to examine bivariate relation-
ships. However, we did construct simple and multivari-
able models of those ED-specific variables that were
available.
In 2007 the survey began asking ED administrators
annually whether they ever boarded patients for more
than 2 hours and whether they used the “full-capacity
protocol” during congested times, that is, the strategy
of moving admitted patients to inpatient hallways or
other parts of the hospital when a hospital bed was not
Pitts et al. • NATIONAL SURVEY OF ED BOARDING
available.13 We analyzed responses to these two ques-
tions using the ED weights provided by NCHS.
In 2009/2010, the survey added the patient visit-level
item “length of time waiting for hospital admission
(minutes) calculated from date and time of bed request
for hospital admissions and date and time patient actu-
ally left the ED.”14 This item serves as our definition of
boarding time, consistent with a recent consensus.15
The ED-level analysis required the NCHS-assigned hos-
pital identifier code, which does not carry over
between years in public use data.16 Therefore, our ED-
level estimates used only the most recent survey year,
2010.14
Data Analysis
We used Stata version 12 for these analyses, including
the suite of survey programs for complex survey analy-
sis when appropriate, using primary sampling unit and
stratum identifiers supplied by NCHS. For each analysis
we used one of two probability weights supplied by
NCHS, one for ED-level analysis and another for visit-
level analysis.
Both the visit-level boarding time and the ED-level
median boarding time were positively skewed, so we
summarized them with the geometric mean rather than
the arithmetic mean. One threat to the validity of this
analysis is a high nonresponse rate for boarding times,
likely due to incomplete ED records rather than errors
in data abstraction. We explored the potential effect of
missing boarding time by replacing the missing board-
ing times with the geometric mean boarding time, mod-
ified by the covariate pattern for that observation.
Missing boarding time was somewhat more common
for ages 15 to 45 years, Hispanic ethnicity, and unin-
sured payer type. We generated predicted boarding
times in multivariable linear regression on the log-
transformed boarding time with age, sex, race, ethnic-
ity, and payer type as independent variables. Of note,
race and ethnicity are also frequently missing and are
themselves routinely imputed for the public use file by
the NCHS. Once these predicted values were generated,
we used the Stata imputation program to sample once
from the distribution of predicted values for each miss-
ing boarding time, to reduce the bias in the standard
error caused by single imputation. The median board-
ing time of the fully imputed sample (76 minutes) was
very close to the observed median (79 minutes), but as
expected with single imputation, the spread of the
imputed distribution was narrower (Table 1).17 Since
boarding time, the only variable with a large missing
fraction, was the dependent variable in most of our
analyses, we did not feel that multiple imputation would
reduce the potential bias caused by missing data, and
we therefore used complete case analysis for all further
analyses.
RESULTS
Between 2007 and 2010, an average of 341 EDs were
included in the sample annually, for a weighted estimate
of 4,760 EDs nationally. Six percent of responses to the
institutional boarding questions were missing. Between
2009 and 2010, a total of 9,117 patient record forms
ACADEMIC EMERGENCY MEDICINE • May 2014, Vol. 21, No. 5 • www.aemj.org 499

Table 1 (mean occupancy per ED treatment space, a proxy for

Characteristics of the Boarding Variable, United States, 2010 ED crowding) could not be calculated because the num-
ber of treatment spaces per ED is not included in the
Nonmissing Missing cases
public use file.
Characteristic cases (n = 6,756) imputed* (n = 9,117) Both unadjusted and adjusted values of median
boarding time are shown in Table 4, with ED character-
Percentiles of the
distribution (in minutes) istics as independent variables, but excluding the char-
10th 10 16 acteristics of ED operations described in Table 3, since
25th 36 45 these values were not available at the ED level. After
50th (median) 79 76 accounting for other terms in this model, the ED-level
75th 145 124
90th 243 213
variables most strongly associated with longer ED
Percent of admissions boarding times were an urban location, being situated
(95% CI) boarding for in the northeast region, and non-Hispanic black ethnic-
0 hours 4 (3–6) † ity and race visit composition. Although there was a
>1 hour 61 (58–65) 63 (60–66) bivariate association between resident/intern encoun-
>2 hours 32 (30–35) 26 (24–29)
>4 hours 10 (9–12) 8 (7–9) ters and boarding, this disappeared when other vari-
>8 hours 3 (2–4) 2 (2–3) ables were included in the multivariable model.

*Model-based single imputation.

†Zeros are undefined in model of ln(boarding time) used for
imputation.
containing the boarding time item were filled out for
patients who were admitted to the hospital, of 69,878
total forms for all visits. This resulted in a weighted esti-
mate of 17.1 million admissions annually. Twenty-four
percent of boarding times were missing, unweighted.
Analysis of the two institutional questions showed
that 61% (95% CI = 54% to 68%) of EDs reported that
they sometimes boarded patients for more than 2 hours,
but only 19% (95% CI = 16% to 22%) used the “full-
capacity” strategy during periods of severe crowding.
In 2010 boarding was reported more frequently in
urban (69%) than rural (34%) EDs and more frequently
in the northeast (71%) than other geographic regions
(53%). EDs reporting boarding saw far more visits (90.7
million patient encounters) than other EDs (30.1 million
patient encounters).
The distribution of boarding times is shown in
Table 1, including percentiles for arbitrary threshold
definitions of boarding. This includes the 2-hour thresh-
old proposed by the 2003 GAO report and the 4-hour
threshold proposed in a recent definition of boarding
by the Joint Commission.18
Of visit characteristics assessed, older age, arrival dur-
ing office hours, arrival by EMS, and advanced imaging
(computed tomography, ultrasound, or magnetic reso-
nance imaging) were associated with longer boarding
duration. Longer boarding durations were not associ-
ated with sex, race/ethnicity, payer type, triage category,
or intensive care unit (ICU) admissions (Table 2). In the
analysis of institutional-level data, among EDs that
reported never boarding patients more than 2 hours
21% of admissions waited in the ED for more than
2 hours for departure after a bed was requested.
Table 3 shows that longer boarding time was associ-
ated with several characteristics of ED operations. EDs
with longer boarding times account for a disproportion-
ate number of patient visits and also had higher propor-
tions of urgent visits, longer waits to be seen, much
higher average occupancies, more hospital admissions,
a higher hospital admission rate, and longer lengths of
stay in the hospital. Individual mean occupancy rates
DISCUSSION
We found that ED boarding of inpatients is common
not only in the many severely overcrowded hospitals
that have driven much current health care policy on the
issue of ED crowding, but also in the aggregate nation-
ally. However, ED boarding is not a universal practice.
Prolonged ED boarding and other throughput intervals
are often viewed as evidence of poor hospital perfor-
mance and have been chosen as quality measures in the
Medicare Hospital Value-based Purchasing Program.
The concern that financial sanctions will have the unin-
tended consequence of penalizing safety net hospitals
was allayed in one recent study of the overall length of
ED visits, also using the NHAMCS survey.19 Our study
similarly finds no effect of payer composition on board-
ing times. However, by aggregating volumes and other
operational data in quartiles of boarding duration, we
were able to infer a relationship between ED volume
and boarding, despite the absence of ED-level annual
volume in the NHAMCS public-use files.
While we identified several individual and ED charac-
teristics that are associated with increased boarding
duration, the most striking finding is the marked step-
up in both ED volume and ED occupancy between the
second and third quartiles of boarding duration noted
in Table 3, implying that rather than increasing continu-
ously with ED volume, boarding becomes a problem
when ED volume and occupancy reach a threshold
value. This finding confirms and refines the results of
another study of NHAMCS overall ED length of visit,20
implying that rather than being a response mainly to
the characteristics of ED patients, boarding is a system
response to a high patient load. The independent asso-
ciation of urbanicity, black race, and northeast region
with boarding duration in our multivariable model may
indeed be confounded or modified in part by ED vol-
ume, since ED volume could not be included in the mul-
tivariable model. This interpretation of our findings
implies that EDs in the United States segregate into
tiers defined by volume thresholds, not safety-net status.
Pay-for-performance in this instance might “steal from
the large and give to the small.”21 It remains to be seen
whether this disparity in boarding time between high-
and low-volume EDs will be influenced favorably by the
500 Pitts et al. • NATIONAL SURVEY OF ED BOARDING

Table 2
Bivariate Analysis of Boarding, by Visit Characteristics, United States, 2009–2010

Percent Percent
Boarding Boarding
More Than More Than
Category (n = Unweighted 2 Hours 2 Hours
Sample Size) (95% CI) p-value* Category (95% CI) p-value
Survey year Triage category (1 = most acute)
2009 (n = 3,334) 32 (29–36) 0.286 1 (n = 229) 31 (22–40) 0.057
2010 (n = 3,422) 33 (29–36) 2 (n = 1,669) 37 (33–41)
3 (n = 3,605) 32 (29–35)
Age category (yr) 4 (n = 856) 29 (23–36)
<20 (n = 519) 23 (18–30) 0.002 5 (n = 182) 25 (17–36)
20–40 (n = 1,107) 29 (25–34)
40–60 (n = 1,842) 34 (31–38) Ambulance arrival
60–80 (n = 2,007) 32 (29–36) No (n = 3,589) 31 (28–34) 0.029
>80 (n = 1,281) 36 (31–40) Yes (n = 2,852) 35 (31–38)
Sex Advanced imaging
Female (n = 3,667) 32 (29–35) 0.745 No (n = 4,229) 30 (27–33) 0.001
Male (n = 3,089) 33 (29–36) Yes (n = 2,527) 36 (32–39)
Race/ethnicity ICU admission
White (n = 4,388) 31 (28–34) 0.127 No (n = 5,294) 33 (30–35) 0.413
Black (n = 1,334) 36 (32–41) Yes (n = 813) 30 (26–36)
Hispanic (n = 704) 32 (26–38)
Other (n = 330) 36 (29–44) Seen by resident or intern
No (n = 5,581) 30 (27–33) 0.001
Payer type 0.416 Yes (n = 1,175) 44 (37–51)
Private (n = 803) 30 (27–35)
Medicare (n = 1,471) 33 (29–39) Do you board outside ED?
Medicaid/CHIP (n = 582) 29 (23–35) No (n = 4,531) 29 (26–33) 0.005
Uninsured (n = 264) 27 (20–36) Yes (n = 1,924) 39 (34–45)
Other (n = 3,636) 33 (30–37)
Do you ever board > 2 hours?
Arrival during office hours No (n = 1,092) 21 (15–28) 0.002
(8 a.m.-5 p.m., M–F) Yes (n = 5,150)
No (n = 4,218) 30 (27–32) <0.001 34 (31–38)
Yes (n = 2,538) 37 (33–41)

ICU = intensive care unit.

*P-values are for the hypothesis of no difference between levels of each category.

Table 3
ED Operational Metrics, Summarized by Quartile of Boarding Time, United States, 2010

Quartile of Median Boarding Time

Metric First Second Third Fourth All EDs
Median boarding time (minutes), range 0–24 25–45 46–86 87–410 0–410
Number of EDs sampled 49 50 110 141 350
Number of EDs, weighted estimate* 1,053 989 1,335 1,345 4,722
Measures related to input
Total number of ED visits (millions) 16.8 18.2 41.3 53.6 129.8
Mean annual visits per ED (visits/EDs, in thousands) 16.0 18.4 30.9 39.9 27.5
Percent high acuity visits† 10.8% 11.8% 10.3% 13.8% 12.0%
Measures related to throughput
Median wait to be seen (minutes) 21 24 30 31 28
90th percentile wait to be seen (minutes) 99 96 113 133 118
Mean ED occupancy (persons present per ED)‡ 5.1 5.8 11.3 16.6 10.3
Measures related to output
Hospital admissions (millions) 1.7 2.0 5.5 8.1 17.2
Admission rate (admissions/visits) 10.1% 10.8% 13.2% 15.1% 13.2%
Median length of hospital stay (days) 3 3 4 4 4

*The NHAMCS oversamples larger EDs. Unequal weights prevented the formation of exactly equal-sized quartiles of boarding.
NHAMCS = National Hospital Ambulatory Medical Care Survey.
†Defined as triage category 1 or 2 in a five-level classification system. Details in NHAMCS on-line documentation.
‡Defined as the weighted sum of lengths of visit in minutes in 2010, divided by the number of minutes in a year.2
ACADEMIC EMERGENCY MEDICINE • May 2014, Vol. 21, No. 5 • www.aemj.org 501

Table 4
Multivariable ED-level Analysis of Boarding Time

Average of Median Boarding Times (minutes)*

Unadjusted Adjusted†
Metric Estimate (95% CI) p-value Estimate (95% CI) p-value
Age category
Pediatric ED 76 (59–97) 0.019 58 (38–90) 0.284
First adult tertile 41 (28–61) 38 (28–51)
Second adult tertile 44 (29–67) 50 (41–61)
Third adult tertile 51 (39–66) 49 (42–58)
Proportion non-Hispanic black race (quartile)
1 (lowest proportion) 31 (22–42) 0.000 35 (28–44) 0.007
2 76 (63–93) 66 (52–84)
3 55 (41–73) 53 (40–69)
4 49 (37–64) 46 (37–58)
Proportion Hispanic (quartile)
1 49 (37, 65) 0.277 57 (46–71) 0.091
2 42 (26, 70) 44 (33–59)
3 39 (26, 59) 38 (30–49)
4 61 (46, 81) 43 (29–63)
Region
Northeast 73 (54–98) 0.049 68 (51–91) 0.029
Midwest 38 (23–62) 43 (34–55)
South 46 (37–58) 45 (38–55)
West 40 (21–74) 36 (23–56)
Poverty quartile‡
1 50 (36–69) 0.372 44 (35–56) 0.938
2 51 (31–85) 50 (37–66)
3 37 (28–49) 46 (39–55)
4 48 (31–75) 44 (32–62)
Urbanicity
Rural 29 (21–40) 0.000 30 (23–39) 0.000
Urban 59 (48–72) 58 (51–66)
Resident/intern encounters
None 38 (29–50) 0.016 43 (36–51) 0.322
1st tertile 56 (42–75) 58 (44–75)
2nd tertile 67 (52–87) 46 (35–62)
3rd tertile 66 (50–87) 44 (32–62)

Operational metrics are excluded from the model due to disclosure risk in public data. United States, 2010.
*Geometric mean. p-values are for the hypothesis of no difference between levels of this category.
†Adjusted for the other variables in this table in linear regression on log-transformed boarding time.
‡Poverty is defined as a Medicaid or uninsured visit.

new performance incentives. This disparity affects more
patients than differences in socioeconomic status,
because larger EDs by definition account for more vis-
its. The half of EDs with the longest median boarding
times in 2010 accounted for 73% of ED visits and 79%
of hospital admissions in our weighted sample.
This negative volume-outcome relationship is at odds
with other clinical outcomes22 and implies an absence
of economies of scale in the ED, when one considers
not just morbidity and mortality, but also the timeliness
of care as an important clinical outcome. Our analysis
supports the development of a more robust operations
management science focusing on system throughput in
high-volume EDs.
The majority of respondents to a Joint Commission
field review wanted a threshold boarding time defined,
to calculate boarding rates.18 Although the Joint Com-
mission’s 4-hour threshold is not being imposed as a
national target, its substitution for the 2-hour threshold
in more common use has a substantial effect on the fre-
quency of boarding, with nationwide boarding rates
dropping from 32% to 10% using the less strict crite-
rion. Tolerance of longer boarding times increases the
unproductive occupancy of ED treatment spaces,
decreases efficiency, and translates into longer patient
waits to be seen. When these thresholds are applied to
the ED-level metric “median boarding time” as defined
in the current Centers for Medicare and Medicaid Ser-
vices core measure, rather than to visit-level boarding
times, the proportion of EDs failing the 2-hour criterion
in our national sample becomes very small indeed, at
9.9% (95% CI = 4.5% to 20%), and the proportion fail-
ing the 4-hour threshold becomes too low for reliable
estimation.
LIMITATIONS
The NHAMCS is a retrospective chart review and thus
is limited in several predictable ways. Reconstructing
throughput measures such as boarding times from the
records of completed ED visits can introduce missing
and inaccurate responses due to inconsistent quality in
a heterogeneous sample of busy EDs with widely differ-
ent documentation styles. The nonresponse rate for the
502
“boarding time” item was 26%, likely due to gaps in
documentation quality rather than abstraction errors.
Although this nonresponse rate is higher than the typi-
cal rates of 5% to 10% for most other NHAMCS items,
this does not invalidate the findings of our analysis.
Recent empirical studies have found nonresponse bias
to be much less important than once believed, and some
have proposed a “rule of thumb” of less than 40% non-
response as a threshold of acceptability.23
We could not evaluate several other potential causes
of boarding because ED identity is excluded from the
public files. This includes hospital occupancy rates, hos-
pital teaching status, and local contextual variables like
primary care capacity and market characteristics.
Although the NCHS provides a hospital code in pub-
lic-use data for ED-level analysis, these files have been
evaluated for disclosure risk and may have been modi-
fied (or “masked”) in unidentified ways to minimize this
risk.12 Finally, the causes of boarding are likely to be
complex, involving interaction between predictive vari-
ables and between levels of analysis (ED-level vs. visit-
level). Both sample size considerations and lack of a
simple method for analyzing multilevel data in complex
samples limit our ability to investigate these complicated
relationships in greater detail. Enterprise-wide data
warehouses are increasingly available and can tally time
stamps on large numbers of consecutive ED visits.
Large simple random samples are easier to analyze
than the multistage, stratified samples required for
NHAMCS, especially in hierarchical models. Larger
sample sizes allow for more nuanced studies of the
causes of boarding, such as interactions between rela-
tively small clinical categories and complaint severity.24
However, these clinical databases are necessarily limited
to their catchment areas and thus are often not suitable
for national policy development.
CONCLUSIONS
Boarding of inpatients in EDs is common nationally,
especially in high-volume hospitals. Because high-vol-
ume EDs board significantly more often than low-vol-
ume EDs, boarding affects a much larger proportion of
patients than hospitals. Our analysis is potentially
biased by missing data, but new reimbursement and
accreditation sanctions may incentivize higher response
rates in the future.
References
1. Kellermann AL, Martinez R. The ER, 50 years on. N
Engl J Med 2011;364:2278–9.
2. Pitts SR, Pines JM, Handrigan MT, Kellermann AL.
National trends in emergency department occu-
pancy, 2001 to 2008: effect of inpatient admissions
versus emergency department practice intensity.
Ann Emerg Med 2012;60:679–86.
3. U.S. Government Accountability Office. Hospital
Emergency Departments: Crowding Continues to
Occur, and Some Patients Wait Longer Than Rec-
ommended Time Frames. Available at: http://www.
gao.gov/new.items/d09347.pdf. Accessed Feb 8,
2014.
Pitts et al. • NATIONAL SURVEY OF ED BOARDING
4. U.S. Government Accountability Office. Hospital
Emergency Departments: Crowded Conditions Vary
Among Hospitals and Communities. Available at:
http://www.gao.gov/new.items/d03460.pdf. Accessed
Feb 8, 2014.
5. Felton BM, Reisdorff EJ, Krone CN, Laskaris GA.
Emergency department overcrowding and inpatient
boarding: a statewide glimpse in time. Acad Emerg
Med 2011;18:1386–91.
6. Ding R, McCarthy ML, Desmond JS, Lee JS, Aron-
sky D, Zeger SL. Characterizing waiting room time,
treatment time, and boarding time in the emergency
department using quantile regression. Acad Emerg
Med 2010;17:813–23.
7. Schneider SM, Gallery ME, Schafermeyer R, Zwe-
mer FL. Emergency department crowding: a point
in time. Ann Emerg Med 2003;42:167–72.
8. Carr BG, Hollander JE, Baxt WG, Datner EM, Pines
JM. Trends in boarding of admitted patients in US
emergency departments 2003–2005. J Emerg Med
2010;34:506–11.
9. U.S. Centers for Medicare and Medicaid Services.
Hospital Compare. Available at: http://www.cms.gov/
Medicare/Quality-Initiatives-Patient-Assessment-Instr-
uments/hospital-value-based-purchasing/index.html.
Accessed Feb 8, 2014.
10. Pitts SR, Carrier ER, Rich EC, Kellermann AL.
Where Americans get acute care: increasingly, it’s
not at their doctor’s office. Health Aff (Millwood)
2010;29:1620–9.
11. Kocher KE, Dimick JB, Nallamothu BJ. Changes in
the source of unscheduled hospitalizations in the
United States. Med Care 2013;51:689–98.
12. Centers for Disease Control and Prevention. Ambu-
latory Health Care Data. Available at: http://www.
cdc.gov/nchs/ahcd.htm. Accessed Feb 8, 2014.
13. Viccellio A, Santora C, Singer AJ, Thode HC Jr,
Henry MC. The association between transfer of
emergency department boarders to inpatient hall-
ways and mortality: a 4-year experience. Ann
Emerg Med 2009;54:487–91.
14. National Center for Health Statistics. 2010 NHAMCS
Micro-data File Documentation. Available at:
ftp://ftp.cdc.gov/pub/Health_statistics/NCHs/Dataset_
Documentation/NHAMCS/doc2010.pdf. Accessed
Feb 8, 2014.
15. Welch SJ, Stone-Griffith S, Asplin B, Davidson SJ,
Augustine J, Schuur JD. Emergency department
operations dictionary: results of the second perfor-
mance measures and benchmarking summit. Acad
Emerg Med 2011;18:539–44.
16. McCaig LF, Burt CW. Understanding and interpret-
ing the national hospital ambulatory medical care
survey: key questions and answers. Ann Emerg
Med 2012;60:716–21.
17. Donders AR, van der Heijden GJ, Stijnen T, Moons
KG. Review: a gentle introduction to imputation of
missing values. J Clin Epidemiol 2006;59:1087–91.
18. The Joint Commission. Patient Flow Through the
Emergency Department. Available at: http://www.
jointcommission.org/assets/1/6/Field_Review_20111201
_EDO_HAP.pdf. Accessed Feb 8, 2014.
ACADEMIC EMERGENCY MEDICINE • May 2014, Vol. 21, No. 5 • www.aemj.org 503

19. Fee C, Burstin H, Maselli JH, Hsia RY. Association 22. Joynt KE, Orav EJ, Jha AK. The association
of emergency department length of stay with safety- between hospital volume and processes, outcomes,
net status. JAMA 2012;307:476–82. and costs of care for congestive heart failure. Ann
20. Pines JM, Decker SL, Hu T. Exogenous predictors Intern Med 2011;154:94–102.
of national performance measures for emergency 23. Johnson TP, Wislar JS. Response rates and nonre-
department crowding. Ann Emerg Med sponse errors in surveys. JAMA 2012;307:1805–6.
2012;60:293–8. 24. McCarthy ML, Zeger SL, Ding R, Aronsky D, Hoot
21. Werner RM. Does pay-for-performance steal from NR, Kelen GD. The challenge of predicting demand
the poor and give to the rich? Ann Intern Med for emergency department services. Acad Emerg
2010;153:340–1. Med 2008;15:337–46.